Deep Convolutional GANs

In this lesson you'll implement the Deep Convolutional GAN model to generate full color images. The additional complexity of these images requires using convolutional layers in the generator and discriminator. You'll also need to use a technique called batch normalization to get the GAN to train appropriately.

You'll be training your GAN on the Street View House Numbers dataset, a small example is shown above. After training, the generator will be able to create images that are nearly identical to these images.